Abstract. Background For the exploration of new anticancer drugs, experiments with tumor-bearing animals are mandatory. However, the results obtained with animals have to be carefully evaluated, since drug metabolisms are quite different between animals and humans (1, 2). The other approach is to use cell cultures that mimic whole body. We have established an in vitro assay method of anti-tumor activity, based on the relative cytotoxicity against human oral squamous cell carcinoma (OSCC) cell lines (Ca9-22, HSC-2, HSC-3, HSC-4) and human normal oral mesenchymal cells (gingival fibroblast, HGF; periodontal ligament fibroblast, HPLF; pulp cell, HPC) (3). The validity of this model for estimating the antitumor activity was obtained by our finding that anticancer drugs such as topoisomerase I inhibitors (camptothecin, SN-38) and topoisomerase II inhibitors (etoposide, teniposide), anthracyclines (doxorubicin, daunorubicin, epirubicin, mitoxantrone), mitomycin C, docetaxel, 5-FU and bacterial products (nocobactin NA-a, -b) showed one or two-orders higher tumor-specificity than hundreds of other natural products including tannins, flavonoids and other lower molecular weight polyphenols (3).We recently found that doxorubicin showed very potent cytotoxicity against human normal oral epithelial cells (mucosa keratinocyte, HOK; primary gingival epithelial cell, HGEP) (4-6). However, the mechanism of keratinocyte toxicity is obscure. In the present study, we investigated whether keratinocyte toxicity is reproducibly observable, regardless the culturing conditions and types of anticancer drugs, and whether it is coupled with apoptosis induction. We also present our preliminary data that searched natural products for their ability to alleviate the keratinocyte toxicity.